Abstract: Accurate air-to-ground channel modeling in rural environments is essential for ensuring the reliability and high-capacity transmission of unmanned aerial vehicle (UAV) communications. Existing models typically rely on limited environmental assumptions and lack multimodal perception of the propagation environment, which restricts their generalization capability under unknown flight trajectories. To address this issue, a rural communication scenario compliant with the 3GPP TR 38.901 standard is established, upon which a multimodal dataset comprising aerial images, depth information, and transceiver parameters is constructed. Based on this dataset, a multimodal spatial perception-based millimeter-wave channel CIR prediction model is proposed. The model is capable of predicting multipath power and delay parameters in complex rural environments and supports flexible prediction for an arbitrary number of multipath components. Experimental results demonstrate that the proposed model achieves high prediction accuracy and strong generalization capability across various rural UAV flight trajectories. This study provides an effective data-driven solution for UAV air-to-ground channel modeling in sparse rural environments.